I present an homogeneous determination of the physical properties of 14 transiting extrasolar planetary systems for which good data are available. The input quantities for each system are the results of the light curve analyses (Paper 1), and published measurements of the stellar velocity amplitude, Teff and [Fe/H]. The physical properties are determined by interpolating within tabulated predictions from stellar theory. Statistical uncertainties are found using a perturbation algorithm, which gives a detailed error budget for every output quantity. Systematic uncertainties are assessed for each quantity by comparing the values found using different stellar models. As a theory-free alternative, physical properties are also calculated using an empirical mass-radius relation constructed using low-mass eclipsing binary stars. The properties of the planets depend mostly on parameters measured from the light and velocity curves, and have only a minor sensitivity to theoretical predictions. In contrast, the orbital semimajor axes and stellar masses have a strong dependence on theoretical predictions, and their systematic uncertainties can be substantially larger than the statistical ones. Using the empirical mass--radius relation instead, their values are smaller by up to 15%. Thus our understanding of extrasolar planets is currently limited by our lack of understanding of low-mass stars. Using all known transiting planets, I find that correlations between their orbital periods, masses and surface gravities are significant at the 2-3 sigma level. However, the separation of the known planets into two classes according to their Safronov number is weaker than previously found, and may not be statistically significant (abridged).